What is CAT Modeling?
Catastrophe (CAT) modeling is the probabilistic simulation of financial losses from natural and man-made catastrophe events — including hurricanes, earthquakes, floods, and wildfires — using stochastic event sets, vulnerability functions, and financial module calculations to estimate expected annual loss (EAL) and probable maximum loss (PML) for insurance portfolios and individual risks.
Catastrophe modeling emerged as a discipline after Hurricane Andrew (1992) and the Northridge earthquake (1994) revealed that traditional actuarial loss estimation techniques — which relied on historical loss experience — were wholly inadequate for quantifying peak natural catastrophe exposures. Modern CAT models replaced historical extrapolation with physics-based simulation: generating tens of thousands of synthetic catastrophe events, estimating their intensity at each insured location, and translating that intensity into expected damage and insured loss.
The major commercial CAT models — developed by Verisk (AIR Worldwide), Moody's Analytics (RMS), and CoreLogic — share a common architecture. The hazard module generates a stochastic event catalogue representing the full range of realistic catastrophe scenarios for a given peril and geography. The vulnerability module translates hazard intensity (wind speed, ground shaking, flood depth) at each location into a mean damage ratio (MDR) and damage distribution, using engineering research and historical claims data. The financial module applies insurance terms — limits, deductibles, coinsurance, reinsurance — to calculate insured loss from the gross damage estimate.
Outputs from CAT models are expressed as exceedance probability (EP) curves: the probability of exceeding various loss levels over a defined time horizon. Key metrics include: Annual Average Loss (AAL or EAL) — the expected loss per year averaged across all scenarios; Probable Maximum Loss (PML) — the loss at a specific return period (e.g., 250-year, 1-in-250 annual probability); and Average Annual Loss as a percentage of TIV (a normalised exposure metric).
Underwriters use CAT models in two ways: at the portfolio level (monitoring aggregate PML relative to reinsurance tower attachment points) and at the risk level (assessing the marginal impact of a new submission on portfolio CAT load). E&S property underwriters writing Gulf Coast wind or California wildfire must understand how each new risk adds to their portfolio's CAT exposure — particularly relative to their reinsurance treaty limits.
CAT model uncertainty is significant: models may disagree on a given risk's AAL by a factor of two or more, and major events regularly produce losses that differ substantially from model expectations. Underwriters must treat model outputs as probabilistic estimates rather than precise predictions.
Orb's CAT and hazard agents surface key modeled metrics — flood zone, wind zone, HVHZ designation, and wildfire hazard score — for every submission address, allowing underwriters to assess the catastrophe dimension of each new risk without running a separate model query. For portfolio-level CAT monitoring, Orb's treaty agent tracks geographic concentration by peril zone as submissions are bound, flagging when portfolio aggregation in a given territory approaches the pre-defined threshold that would trigger reinsurance treaty constraints.
Frequently asked questions
What is a Probable Maximum Loss (PML)?
PML is the estimated maximum loss from a single catastrophic event at a specified return period — typically expressed as a percentage of total insured value (TIV) or as a dollar amount. A 250-year PML (0.4% annual exceedance probability) represents the loss level that would be equalled or exceeded in the worst event you might expect to experience over a 250-year period. Reinsurance towers are typically structured so that PML at a chosen return period is covered by the reinsurance program.
How often are CAT models updated?
Major CAT model vendors typically release significant model updates every 2–5 years, with interim updates for specific perils or regions. After major events — such as Hurricane Ian (2022) or the California wildfires — model vendors often release revised versions incorporating new data on event characteristics, vulnerability, and claims experience. Model updates can result in significant changes to estimated losses for the same portfolio.
What is the difference between AIR and RMS CAT models?
AIR Worldwide (Verisk) and RMS (Moody's Analytics) are the two dominant commercial CAT model vendors. Both cover the same major perils and geographies but use different modelling methodologies, event catalogues, and vulnerability functions — resulting in different loss estimates for the same portfolio. Many insurers and reinsurers run both models and use a blend or the more conservative estimate for pricing and capital purposes.